Judging Destinations by Structure, Evidence & Friction
Commercial search engines reward SEO preambles, affiliate roundup mills, and pages packed with display ad networks. Crank retrieves candidate URLs from upstream recall and re-ranks them using deterministic HTML and structured-data extractors:
Recipes (recipes.v0.1): Parses schema.org/Recipe and DOM structure to reward clear ingredient lists, metric gram weights, and first-hand testing notes while penalizing 1,000-word SEO preambles, ad scripts, and modal popups.
Product Research (products.v1): Elevates stated testing methodology, lab measurements, hardware teardowns, diagnostic/repair guides, and enthusiast community threads while penalizing single-brand e-commerce catalogs and high-affiliate roundup listicles.
PageYank Lab Friction: Joins real-browser friction measurements (domain_yank) to demote pages with severe layout shifts, autoplay video players, and tracker bloat.
Missing Evidence Is Never Treated as Evidence of Absence
Every result card includes a “Why this page” inspection drawer showing the exact linear formula and measured attributes that determined its position.
When a candidate page has not been fetched within the live latency budget (or blocks automated inspection), Crank never invents a fallback penalty or claims “0 ad networks.” Uninspected pages are explicitly marked Not assessed (unfetched) and ranked neutrally by query relevance and any known domain-level PageYank measurements.
What Crank Stores and What Leaves the Server
No Behavioral Tracking: Crank Search uses zero third-party analytics SDKs, zero tracking cookies, and zero ad networks.
Outgoing Link Protection: All Crank pages send Referrer-Policy: no-referrer and mark external links with rel="noopener noreferrer" so destination websites never see the search query you typed.
Crawler Protection: Dynamic search result routes (/search?q=...) send X-Robots-Tag: noindex, nofollow so queries are never indexed by third-party search crawlers.
Upstream Recall Provider: When a query is not already cached in Crank's database, the search phrase is forwarded server-side to the Brave Search API to retrieve candidate URLs without forwarding your IP address, cookies, or browser headers.